使用最长公共子序列计算文本相似度

这里一个比较合理的思路。
如果最长公共子序列长为k， 文本1长为n1， 文本2长为n2，则相似度为 (2 * k) / (n1 + n2)

结果如下：

代码如下：

#include <iostream>
#include <string>
#include <windows.h>
#include <algorithm>
#include <cmath>
#include <vector>
#include <tchar.h>

using namespace std;

class Matcher{
public:
	Matcher(wchar_t s1[], int n1, wchar_t s2[], int n2){
		this->s1 = s1;
		this->s2 = s2;
		this->n1 = n1;
		this->n2 = n2;
		buf = new int*[n1 + 1];
		for(int i = 0; i <= n1; i++){
			buf[i] = new int[n2 + 1];
			for(int j = 0; j <= n2; j++){
				buf[i][j] = 0;
			}
		}
	}

	~Matcher(){
		for(int i = 0; i <= n1; i++){
			delete[] buf[i];
		}
		delete[] buf;
	}

	/*计算相似度*/
	double match(){
		//先排个序
		sort(s1, s1 + n1);
		sort(s2, s2 + n2);

		int maxCommonLen = matchAct(n1, n2);
		return 2 * maxCommonLen * 1.0 / (n1 + n2);
	}

private:
	wchar_t* s1;
	wchar_t* s2;
	int n1;
	int n2;
	int** buf;

	/*查找最长公共子序列长度*/
	int matchAct(int n1, int n2){
		if(n1 <= 0 || n2 <= 0){
			return 0;
		}

		//Dynamic Planning--如果去掉以下三个if，则是普通的分治
		if(buf[n1 - 1][n2 - 1] == 0){
			buf[n1 - 1][n2 - 1] = matchAct(n1 - 1, n2 - 1);
		}

		if(buf[n1 - 1][n2] == 0){
			buf[n1 - 1][n2] = matchAct(n1 - 1, n2);
		}

		if(buf[n1][n2 - 1] == 0){
			buf[n1][n2 - 1] = matchAct(n1, n2 - 1);
		}

		if(s1[n1 - 1] == s2[n2 - 1]){
			return buf[n1 - 1][n2 - 1] + 1;
		}

		return max(buf[n1 - 1][n2],
				buf[n1][n2 - 1]);
	}
};


int main()
{
	wchar_t s1[] = L"a1b2c3d4";
	wchar_t s2[] = L"1452364";
	Matcher m1(s1, 8,  s2, 7);
	cout << "相似度:" << m1.match() << endl;

	wchar_t s3[] = L"我和我的祖国";
	wchar_t s4[] = L"国祖的我和我";
	Matcher m2(s3, 6,  s4, 6);
	cout << "相似度:" << m2.match() << endl;


	wchar_t s5[] = L"东方甲乙木";
	wchar_t s6[] = L"南方丙丁火";
	Matcher m3(s5, 5,  s6, 5);
	cout << "相似度:" << m3.match() << endl;

	return 0;
}